JP2020060981A - Node search method and node search program - Google Patents

Abstract

To identify data portions relating to other data sets, in a data set described based on a plurality of ontologies.SOLUTION: A computer identifies, from nodes included in a first data set described based on a plurality of ontologies, nodes corresponding to each of a plurality of data included in a second data set (step 701). The computer determines, from the identified nodes on a path tracing links included in the first data set, a relationship between the ontologies of a moving source node and a moving destination node (step 702). The computer searches, by tracing a link between the moving source node and the moving destination node based on the determination result, a common node where a first path tracing the link from a first node intersects a second path tracing the link from a second node, (step 703). The computer outputs the search result indicating the common node, the first node, and the second node (step 704).SELECTED DRAWING: Figure 7

Description

  The present invention relates to a node search method and a node search program.

  In recent years, various data sets are described by RDF (Resource Description Framework), and different data sets are linked by RDF. Therefore, different data sets can be combined and analyzed.

  In RDF, three elements of a subject, a predicate, and an object are used as a minimum unit, and this minimum unit is called a triple. For example, in the case of the triple (FJ company, industry, electrical equipment), the subject is “FJ corporation”, the predicate is “business industry”, and the object is “electric equipment”. This triple describes the information that "the industry of FJ company is electrical equipment".

  The subject and the predicate are expressed by a URI (Uniform Resource Identifier), and the object is expressed by a URI or a literal (character string). The URI is described by enclosing it in <>, and the literal is described by enclosing it in "". A URL (Uniform Resource Locator) may be used as the URI. Predicates may be referred to as attributes or properties, and objects may be referred to as attribute values or property values.

  The dataset described by RDF is a set of triples. When describing a dataset as a graph, the subject and object of triples are called nodes and the predicates are called links or edges.

  FIG. 1 shows an example of a data set described by RDF. A value that starts with a combination of an underscore "_" and a colon ":" like _: f represents an arbitrary URI. _: F, "FJ company", "Kawasaki City, Kanagawa", and "electrical device" represent nodes, and <Name>, <Location>, and <Industry> arrows represent links. The arrow of the link indicates the reference direction, and the node at the starting point of the arrow refers to the node at the ending point of the arrow by the link. This data set shows that the name of a certain company is “FJ Company”, the location is “Kawasaki City, Kanagawa Prefecture”, and the industry is “electric equipment”.

The data set in Fig. 1 is described as follows in N-Triples format.
(_: F, <name>, “FJ company”)
(_: F, <Location>, “Kawasaki City, Kanagawa”)
(_: F, <industry>, “electrical equipment”)

When the data set in FIG. 1 is described in Turtle format, it becomes as follows.
_: F <name> “FJ company”;
<Location> “Kawasaki City, Kanagawa Prefecture”;
<Industry> “Electrical equipment”.

  In RDF, information is described according to some ontology definition. For example, when describing a corporation in RDF, an ontology about the corporation is used. There may be multiple ontologies for the same information. For example, there are Organization ontologies, common vocabulary bases, etc. as ontologies related to corporations. Information described based on different ontologies is described by using different structures and links even if the information is the same.

  FIG. 2 shows an example of a data set described based on different ontologies. FIG. 2A shows an example of a data set described based on the Organization ontology. _: F, ●, “FJ company”, and “Kanagawa prefecture” represent nodes, and arrows of <label>, <registered address>, <place>, and <region> represent links. ● represents a blank node. This data set indicates that the name of a certain company is "FJ company" and the prefecture to which the address belongs is "Kanagawa prefecture".

  FIG. 2B shows an example of a data set described based on the common vocabulary base. _: F, ●, “FJ company”, and “Kanagawa prefecture” represent nodes, and <Name>, <notation>, <address>, and <prefecture> arrows represent links. This data set describes the same information as the data set of FIG.

  A single data set may be described by combining multiple ontologies, and a plurality of data sets may be linked by a link to generate a composite data set.

  FIG. 3 shows an example of a composite data set. The data set 311 is described based on the ontology 321 to the ontology 324, and the data set 312 is described based on the ontology 321, the ontology 325, and the ontology 326. Then, the data set 301 is generated by concatenating a plurality of data sets including the data set 311 and the data set 312.

  Further, among the ontologies, there are special ontologies used to describe only specific data sets, and there are basic ontologies used to describe various data sets.

  Regarding RDF, a similarity calculator for calculating similarity between drugs using open data, a computer for integrating non-conceptual data items into a data graph, and a method for obtaining hierarchical information of flat data are known. (See, for example, Patent Documents 1 to 3).

JP, 2016-212853, A JP, 2016-15124, A JP 2012-141955 A

  As described above, multiple datasets described by RDF can be linked by links, and the datasets can be combined and analyzed.

  However, not all datasets are described in RDF. For example, a list of customer companies including attributes such as official names, addresses, and telephone numbers of companies is described in CSV (Comma-Separated Values), and open data of each company may be described in RDF. In this case, the correspondence relationship between the customer company list described in CSV and the data set described in RDF is unknown, and it is difficult to analyze these data in combination.

  Note that such a problem is not limited to the case of associating the data set described by RDF and the data set described by CSV, but in the case of associating the data set described based on a plurality of ontologies with other data sets. Also occurs.

  In one aspect, the present invention aims to identify a data portion related to other data sets in a data set described based on a plurality of ontologies.

In one proposal, the computer performs the following processing.
(1) The computer identifies a node corresponding to each of the plurality of data included in the second data set from the nodes included in the first data set described based on the plurality of ontology.
(2) The computer determines the relevance between the ontology of the source node and the ontology of the destination node on the route that follows the link included in the first data set from each of the identified plurality of nodes.
(3) The computer searches for a common node where the first route and the second route intersect by tracing the link between the source node and the destination node based on the result of determining the relevance. The first route is a route that follows the link from the first node among the plurality of identified nodes, and the second route is a route that follows the link from the second node among the plurality of identified nodes.
(4) The computer outputs a search result including information indicating the common node, information indicating the first node, and information indicating the second node.

  According to the embodiment, in a data set described based on a plurality of ontologies, a data part related to another data set can be specified.

It is a figure which shows the data set described by RDF. It is a figure which shows the data set described based on a different ontology. It is a figure which shows a composite data set. It is a figure which shows the process which matches a customer company list and open data of a company. It is a figure which shows the search process in a composite RDF data set. It is a functional block diagram of a node search device. It is a flowchart of a node search process. It is a functional block diagram which shows the specific example of a node search apparatus. It is a figure which shows the RDF data set contained in a RDF data set group. It is a figure which shows the RDF data set used for calculation of PMI (x, y). It is a figure which shows the RDF data set used for calculation of tfidf (i, j). It is a flowchart which shows the specific example of a node search process. It is a flowchart of a node detection process. It is a figure which shows a search node queue. It is a flowchart of a node movement process. It is a figure which shows a route. It is a figure which shows a route list. It is a figure which shows the route list containing a common node. It is a block diagram of an information processing apparatus.

Hereinafter, embodiments will be described in detail with reference to the drawings.
Hereinafter, a dataset described by RDF may be referred to as an RDF dataset, and a dataset not described by RDF may be referred to as a non-RDF dataset.

  FIG. 4 shows an example of a process of associating a list of customer companies with open data of companies. FIG. 4A shows an example of a customer company list described in CSV. In this example, the name, address, and telephone number of FJ company are recorded in the customer company list.

  FIG. 4B shows an example of open data of a company described by RDF. _: F, _: n1, _: n2, "FJ company", "Kanagawa prefecture Kawasaki city", and "aaa-bbb-ccc" represent nodes, <name>, <head office>, <location>, And <telephone number> arrows represent links. This RDF data set indicates that the name of the company is “FJ company”, the location of the head office is “Kawasaki City, Kanagawa Prefecture”, and the telephone number of the head office is “aaaa-bbb-ccc”.

  FIG. 4C shows an example of the result of associating the customer company list of FIG. 4A with the open data of FIG. 4B. The corresponding node represents a node having a literal corresponding to the data included in the customer company list in the RDF data set, and the common node represents a node that directly or indirectly refers to two or more corresponding nodes.

  When the common node directly refers to the corresponding node, there is one link from the common node to the corresponding node, and when the common node indirectly refers to the corresponding node, the common node goes through one or more other nodes. Then, there are two or more links to the corresponding node.

  For example, the common node “_: n1” directly refers to the corresponding node “Kawasaki City, Kanagawa Prefecture” by the link <Location>, and the corresponding node “aaaa-bbb-ccc” via the node “_: n2”. "Refers to indirectly.

  On the other hand, the common node “_: f” directly refers to the corresponding node “FJ company” by the link <name>, and indirectly connects to the corresponding node “Kawasaki City, Kanagawa Prefecture” via the node “_: n1”. Refer to. Further, the common node “_: f” indirectly refers to the corresponding node “aaa-bbb-ccc” via the node “_: n1” and the node “_: n2”.

  The computer determines whether the character string of each data included in the customer company list of FIG. 4A matches the literal character string of each node included in the RDF data set of FIG. 4B. By checking, the corresponding node can be detected. As a result, three corresponding nodes of “FJ company”, “Kawasaki city, Kanagawa prefecture”, and “aaa-bbb-ccc” are detected from the RDF data set.

Next, the computer can detect the common node from the RDF data set by recursively performing a search process of following a link from each corresponding node in the direction opposite to the reference direction until there are no links that can be followed. In the search process, when a route that follows a link intersects from each of a plurality of corresponding nodes, the node corresponding to the intersection of the routes is detected as a common node. For example, the computer performs the search process in the following procedure.
(P1) The computer traces the link <name> referring to “FJ company” in the reverse direction on the path starting from the corresponding node “FJ company”, and detects the node '_: f'.
(P2) The computer follows the link <location> referring to “Kawasaki City, Kanagawa” in the reverse direction on the route starting from the corresponding node “Kawasaki City, Kanagawa”, and detects the node '_: n1'. .
(P3) The computer follows the link <telephone number> referring to “aaa-bbb-ccc” in the reverse direction on the route starting from the corresponding node “aaa-bbb-ccc”, and the node '_: n2' To detect.
(P4) The computer detects the node '_: f' by tracing the link <head office> referring to '_: n1' in the reverse direction on the route starting from the corresponding node “Kawasaki, Kanagawa”.
(P5) The computer detects the node '_: n1' by tracing the link referring to '_: n2' in the reverse direction on the route starting from the corresponding node'aaa-bbb-ccc '.
(P6) The computer follows the link <headquarters> referring to “_: n1” in the reverse direction on the route starting from the corresponding node “aaa-bbb-ccc”, and detects the node “_: f”. .

  In this case, since the three routes starting from "FJ company", "Kawasaki city, Kanagawa prefecture", and "aaa-bbb-ccc" intersect at node "_: f", node "_: f" becomes Detected as a common node. Also, since two routes starting from "Kawasaki City, Kanagawa" and "aaa-bbb-ccc" intersect at node "_: n1", node "_: n1" is also detected as a common node.

  In this way, by specifying the corresponding node and the common node in the RDF data set, the range of nodes directly or indirectly referenced from the common node can be specified as the data portion related to the non-RDF data set. . This allows non-RDF datasets to be associated with RDF datasets and allows these datasets to be combined and analyzed.

  As an example, in the RDF data set of FIG. 4B, the node '_: f', the node '_: n1', or the node '_: n2' refers to another node by another link (not shown). Suppose that. The information held by another node may be information about a sales office, a business office, a factory, etc., which is not included in the customer company list. In this case, the computer can acquire and analyze information possessed by another node by following another link in the reference direction.

  By the way, in the composite RDF data set shown in FIG. 3, among all the connected data sets, the data portion related to the non-RDF data set is often at most one or two data sets. . In this case, if all connected data sets are searched, irrelevant data sets are also searched, resulting in enormous amount of time calculation and space calculation.

  FIG. 5 shows an example of a search process in a composite RDF dataset including the RDF dataset of FIG. In the RDF data set of FIG. 5, information of a company DB (database) and an abbreviation DB is mixed. "Abbreviation DB", "FJ", and "F" represent nodes, and an arrow of <abbreviation> represents a link.

In this case, the computer performs the search process in the following procedure in order to specify the data portion related to the customer company list in FIG.
(P11) The computer follows the link <formal name> referring to “FJ company” in the reverse direction on the first path starting from the corresponding node “FJ company”, and detects the blank node ●.
(P12) The computer follows the link <name> referring to “FJ company” in the opposite direction on the second path starting from the corresponding node “FJ company” to detect the node “_: f”.
(P13) The computer follows the link <location> referring to “Kanagawa Prefecture Kawasaki City” in the reverse direction on the route starting from the corresponding node “Kanagawa Prefecture Kawasaki City”, and detects the node '_: n1'. .
(P14) The computer follows the link <telephone number> referring to “aaa-bbb-ccc” in the reverse direction on the route starting from the corresponding node “aaa-bbb-ccc”, and the node '_: n2' To detect.
(P15) The computer follows the link referring to ● in the reverse direction on the first route starting from the corresponding node “FJ company”.
(P16) The computer detects the node '_: f' by tracing the link <head office> referring to '_: n1' in the reverse direction on the route starting from the corresponding node “Kawasaki, Kanagawa”.
(P17) The computer detects the node '_: n1' by tracing the link referring to '_: n2' in the reverse direction on the route starting from the corresponding node'aaa-bbb-ccc '.
(P18) The computer follows the link next to the link referring to ● in the reverse direction on the first route starting from the corresponding node “FJ company”.
(P19) The computer follows the link <head office> that refers to “_: n1” in the opposite direction on the route starting from the corresponding node “aaa-bbb-ccc”, and detects the node “_: f”. .
(P20) The computer further follows the next link in the reverse direction on the first route starting from the corresponding node “FJ company”.

  Similar to the case of the RDF data set of FIG. 4B, the range of nodes directly or indirectly referenced from the common node '_: f' is the data portion related to the customer company list. Therefore, originally, the processing of the procedure (P11), the procedure (P15), the procedure (P18), and the procedure (P20) of following the link on the route including the blank node ● is unnecessary. In particular, when a route including the blank node ● includes many links, the amount of calculation for tracing those links becomes enormous. In the search process, if only the data set to which the corresponding node belongs can be searched, it becomes unnecessary to search an irrelevant data set.

  As shown in FIG. 3, each data set is described based on a specific set of ontology. Therefore, a method is conceivable in which a data set to which the corresponding node belongs is obtained, a set of ontology frequently used in the obtained data set is obtained, and only nodes belonging to the set of ontology are selected as search targets. However, it is difficult to identify the set of frequently used ontologies in each data set.

  FIG. 6 illustrates a functional configuration example of the node search device according to the embodiment. The node search device 601 of FIG. 6 includes a storage unit 611, a search unit 612, and an output unit 613. The storage unit 611 stores a first data set 621 and a second data set 622 described based on a plurality of ontologies. The search unit 612 uses the first data set 621 and the second data set 622 to perform node search processing.

  FIG. 7 is a flowchart showing an example of the node search processing performed by the node search device 601 of FIG. First, the search unit 612 identifies a node corresponding to each of the plurality of data included in the second data set 622 from the nodes included in the first data set 621 (step 701).

  Next, the search unit 612 determines the relevance between the ontology of the source node and the ontology of the destination node on the route that follows the links included in the first data set 621 from each of the identified nodes. (Step 702).

  Next, the search unit 612 searches for a common node where the first route and the second route intersect by tracing the link between the source node and the destination node based on the result of determining the relevance. (Step 703). The first route is a route that follows the link from the first node among the plurality of identified nodes, and the second route is a route that follows the link from the second node among the plurality of identified nodes.

  Next, the output unit 613 outputs a search result including information indicating the common node, information indicating the first node, and information indicating the second node (step 704).

  According to the node search device 601 of FIG. 6, it is possible to specify a data portion related to another data set in the data set described based on a plurality of ontologies.

  FIG. 8 shows a specific example of the node search device 601 of FIG. The node search device 801 in FIG. 8 includes a storage unit 811, a calculation unit 812, a search unit 813, and an output unit 814. The storage unit 811, the search unit 813, and the output unit 814 correspond to the storage unit 611, the search unit 612, and the output unit 613 of FIG. 6, respectively.

  The storage unit 811 stores the RDF data set group 821, the ontology group 822, the RDF data set 825, and the non-RDF data set 826. The RDF data set 825 and the non-RDF data set 826 correspond to the first data set 621 and the second data set 622 of FIG. 6, respectively.

  The non-RDF data set 826 may be a data set described in CSV, TSV (Tab-Separated Values), SSV (Space-Separated Values), XML (Extensible Markup Language), a relational data model, or the like.

  For example, the non-RDF dataset 826 may be financial information of a lender company held by a bank, and the RDF dataset 825 may be open data of the firm. In this case, by associating these data sets, it is possible to comprehensively analyze the financial information and open data of the loan recipient company and determine whether or not the bank can lend.

  Further, the non-RDF data set 826 may be personal information of a company officer, and the RDF data set 825 may be open data of the company. In this case, by associating these data sets, it becomes possible to comprehensively analyze the personal information of the officer and the open data of the company, and determine whether or not the transaction with the company is possible.

  The RDF data set group 821 is a plurality of RDF data sets used for calculation of statistics regarding links, and the ontology group 822 is a plurality of ontology for describing those RDF data sets. Each link included in the RDF data set group 821 is defined by each ontology included in the ontology group 822.

  The calculation unit 812 obtains the number of appearances of the links simultaneously appearing for the combination of two links included in each RDF data set of the RDF data set group 821. Then, the calculation unit 812 calculates the co-occurrence statistic 823 for those links using the obtained number of appearances, and stores the co-occurrence statistic 823 in the storage unit 811.

  The calculation unit 812 also extracts words included in the label and comment of each link from each ontology included in the ontology group 822, and obtains the number of appearances of each word that appears in the label and comment of each link. Then, the calculating unit 812 calculates the importance statistic 824 indicating the importance of each word using the obtained number of appearances, and stores the importance statistic 824 in the storage unit 811.

  The search unit 813 identifies a corresponding node corresponding to each of the plurality of data included in the non-RDF data set 826 from the nodes included in the RDF data set 825. Then, the search unit 813 generates a search node queue 827 including the identified corresponding nodes, and stores the search node queue 827 in the storage unit 811. The node registered in the search node queue 827 is used as a movement source node.

  Next, the search unit 813 determines, from each of the plurality of corresponding nodes included in the search node queue 827, the source node and the destination node on the route that follows the link included in the RDF data set 825 in the direction opposite to the reference direction. Determine the relationships between the ontologies of. At this time, the search unit 813 determines the relevance between the ontologies based on the co-occurrence statistic 823 and the importance statistic 824.

  When the search unit 813 determines that there is a relationship between the ontologies, the search unit 813 follows the link between the source node and the destination node in the reverse direction, and sets the destination node in the search node queue 827 as a new source node. Set. Then, the search unit 813 continues the search on the route including the new source node.

  The search unit 813 generates a route list 828 including routes that follow the links included in the RDF data set 825 from each of the corresponding nodes, and stores the route list 828 in the storage unit 811. Then, the search unit 813 continues the search while updating the search node queue 827 and the route list 828, so that the first route following the link from the first corresponding node and the second route following the link from the second corresponding node. Search for a common node where and intersect.

  If the search unit 813 determines that there is no relationship between the ontologies during the search, the search unit 813 terminates the search on the route including the source node. Then, the search unit 813 generates a search result 829 including information indicating the searched common node, information indicating the first node, and information indicating the second node, and stores the search result 829 in the storage unit 811. The output unit 814 outputs the search result 829.

  An ontology is considered to be information that defines what kind of link is used in what kind of structure. Therefore, the calculation unit 812 calculates the co-occurrence statistic 823 and the importance statistic 824 by using the link and the relationship between the links, instead of directly handling the ontology and the relationship between the ontologies. This makes it possible to identify a set of ontologies that can be used together in the same data set based on the co-occurrence statistic 823 and the importance statistic 824.

  When following a link on a route, the search unit 813 may use the source node ontology and the destination node ontology together in the same data set based on the co-occurrence statistic 823 and the importance statistic 824. Estimate whether it is possible or not. Then, the search unit 813 updates the routes in the route list 828 by following the links between the nodes only when the ontology can be used together.

  According to the node search device 801 in FIG. 8, by specifying the corresponding node and the common node in the RDF data set 825, the range from the common node to the corresponding node is specified as the data portion related to the non-RDF data set 826. be able to. As a result, it is possible to acquire data other than the related data part that is referred to by the common node.

  As an example, it is assumed that the common node directly or indirectly refers to another node other than the corresponding node in the RDF dataset 825, and the other node has information that is not included in the non-RDF dataset 826. To do. In this case, by following the link from the common node in the reference direction, it becomes possible to acquire and analyze the information possessed by another node.

  Further, in the search on the route, when it is determined that there is no relationship between the ontology of the source node and the ontology of the destination node, the search on the route is aborted to search for a data portion unrelated to the non-RDF data set 826. Is omitted. As a result, the amount of calculation for associating the RDF data set 825 and the non-RDF data set 826 can be reduced.

  For example, in the RDF data set of FIG. 5, when it is determined that there is no relationship between the ontology of the corresponding node “FJ company” and the blank node ●, the search on the route including those nodes is terminated. As a result, the process of following the links on the route including the blank node ● is omitted, so that the amount of calculation for associating the RDF data set and the customer company list is reduced.

  FIG. 9 shows an example of RDF data sets included in the RDF data set group 821. _: B, _: a, 14a: 100001100xxxx, "ABC", "ABC Library", "100001100xxxx", "national institution", and "new" represent nodes. The arrows of dbo: aaa, dbo: nnn, skos: ppp, org: iii, dct: sss, and org: ccc represent links.

  In the Turtle format of RDF, “dbo:”, “skos:”, “org:”, and “dct:” are called prefixes and are used to omit the beginning part of the URI. By convention, the same prefix is used for the same ontology, so in the example of FIG. 9, the ontologies can be distinguished by the prefix. Therefore, it can be seen that the RDF data set of FIG. 9 is described based on four ontologies indicated by “dbo:”, “skos:”, “org:”, and “dct:”.

  As the co-occurrence statistic 823, for example, the self-mutual information (PMI) of the following equation can be used.

  In Expression (1), PMI (x, y) represents the probability that the link x and the link y co-occur, p (x) represents the appearance probability that the link x appears, and p (y) is the link It represents the appearance probability that y appears. c (x) represents the number of appearances of the link x, c (y) represents the number of appearances of the link y, and c (x, y) represents the number of appearances of the link x and the link y at the same time. Indicates the number of times. N represents the number of appearances of all links, and K represents the number of appearances of all combinations (all co-occurrence) of two links.

  First, the calculation unit 812 calculates c (x), c (y), and N using all RDF data sets included in the RDF data set group 821. Next, the calculation unit 812 obtains c (x, y) using the single RDF data set including the link x and the link y. Then, the calculation unit 812 calculates PMI (x, y) using the equation (1).

  FIG. 10 shows an example of the RDF data set used for the calculation of PMI (x, y). FIG. 10A shows an example of the RDF data set including the node n01. This RDF data set includes nodes n01 to n07 and links L1 to L3. FIG.10 (b) has shown the example of the RDF data set containing the node n11. This RDF data set includes nodes n11 to n17, link L2, link L4, and link L5.

In this case, the number of appearances of each link and the number of appearances of two links appearing at the same time are as follows.
c (L1) = 2
c (L2) = 4
c (L3) = 2
c (L4) = 2
c (L5) = 2
c (L1, L2) = 4
c (L1, L3) = 4
c (L1, L4) = 0
c (L1, L5) = 0
c (L2, L3) = 4
c (L2, L4) = 4
c (L2, L5) = 4
c (L3, L4) = 0
c (L3, L5) = 0
c (L4, L5) = 4

  Since N = 12, when logarithm with base 10 is used as the log of equation (1), P (L1, L2), P (L1, L3), and P (L1, L4) are Calculated as the formula.

  NaN of Formula (4) represents a non-number. PMI (x, y) for other combinations of links are also calculated in the same manner as equations (2) to (4).

  The greater the number of appearances of the link x and the link y at the same time in each RDF data set, the larger the PMI (x, y) of the equation (1). Therefore, the larger PMI (x, y), the more likely the ontology defining link x and the ontology defining link y are used together in the same RDF dataset. Therefore, PMI (x, y) can be utilized to determine the relevance between those ontologies.

  As the importance statistic 824, for example, tfidf (i, j) in the following equation can be used.

は、リンクｄ（ｊ）におけるすべての単語の出現回数の和を表す。式（１３）の｜Ｄ｜は、ＲＤＦデータセット群８２１に含まれる、異なるリンクの個数を表し、｜｛ｄ：ｄ∋ｔ（ｉ）｝｜は、単語ｔ（ｉ）を含むリンクの個数を表す。 N (i, j) in the equation (12) represents the number of appearances of the word t (i) in the link d (j),

Represents the sum of the number of appearances of all words in the link d (j). In Expression (13), | D | represents the number of different links included in the RDF dataset group 821, and | {d: d∋t (i)} | is the number of links including the word t (i). Represents

  First, the calculation unit 812 extracts the links d (j) from all the RDF datasets included in the RDF dataset group 821, and selects the ontology defining the links d (j) from the ontology group 822. Next, the calculation unit 812 extracts words included in the label and comment of the link d (j) from the selected ontology, and obtains the number of appearances of the word t (i) that appears in the extracted label and comment. Then, the calculation unit 812 calculates tfidf (i, j) from Expressions (11) to (13) using the obtained number of appearances as n (i, j).

FIG. 11 shows an example of the RDF data set used to calculate tfidf (i, j). l4a: 100001100xxxx, "ABC Library", "100001100xxxx", and "new" represent nodes, and the arrows of skos: ppp, org: iii, and org: ccc represent links. As the label and comment of each link, for example, the following texts are extracted.
skos: ppp
label: ppp lll
comment: skos: ppp, skos: aaa and skos: hhh are p1 d1 p2. T12 r3 of skos: ppp is t12 c4 of R5 p6 l7.A r8 h13 no m14 t15 one v9 of skos: ppp p16 l10 t11.

org: iii
label: iii
comment: G1 an iii, s28 as a c2 r3 n4, t29 c30 be u31 to u32 to u5 i6 the organization.M7 d7 n9 and i10 iii s11 are a12.The o13 o14 is n15 to w33 s16 are u36. The p17 iii s11 s37 be i18 by the d19 of the iii v20.U38 d19 to d21 the n22 s11 u39 is c23 w40 r24 b25 p25 for'skos: nnn 'of w34 t35 p26 is a s27.

org: ccc
label: ccc bbb
comment: I1 a c2 event w3 r4 in a c5 to t6 organization.D7 on the event the organization may or may not h12 c8 to e9 a10 the event.I11 of 'org: ooo'.

  However, the label and the comment are virtual texts for explaining the embodiment.

  If the RDF dataset group 821 includes only the RDF dataset of FIG. 11, the link d (j) is org: ccc, and the word t (i) is “organization”, then n (i, j). Is 2. Also, since the word included in the org: ccc label is two and the word included in the comment is 32, the sum of the number of appearances of all words is 34.

  The number of different links included in the RDF data set of FIG. 11 is three, and the number of links including “organization” in the label and comment is org: iii and org: ccc. Therefore, when natural logarithm is used as the log of the equation (13), tfidf (i, j) is calculated by the following equation.

  Further, when the link d (j) is org: ccc and the word t (i) is “event”, n (i, j) is 3, and a link including “event” in the label and the comment is There is only one org: ccc. In this case, tfidf (i, j) is calculated by the following equation.

  The tfidf (i, j) for other link and word combinations are also calculated in the same manner as the equations (14) and (15).

  The larger tfidf (i, j) in Expression (11), the greater the importance of the word t (i) in the link d (j). Thus, if the same word tfidf (i, j) shows a large value in two links, then it is likely that the two ontologies that define those links will be used together in the same RDF dataset. Therefore, tfidf (i, j) can be used to determine the relevance between those ontologies.

  Next, the operation of the node search device 801 in FIG. 8 will be described in more detail with reference to FIGS. 12 to 18.

  FIG. 12 is a flowchart showing a specific example of the node search processing performed by the node search device 801 of FIG. First, the calculation unit 812 calculates the co-occurrence statistic 823 for a combination of two links included in each RDF data set of the RDF data set group 821 (step 1201).

  Next, the calculation unit 812 extracts the words included in the label and the comment of each link included in the RDF data set group 821 and calculates the importance statistic 824 for each word (step 1202). Then, the search unit 813 uses the co-occurrence statistic 823 and the importance statistic 824 to perform node detection processing (step 1203).

  FIG. 13 is a flowchart showing an example of the node detection processing in step 1203 of FIG. First, the search unit 813 identifies the corresponding node corresponding to each of the plurality of data included in the non-RDF data set 826 from the nodes included in the RDF data set 825 (step 1301).

  At this time, the search unit 813 checks whether or not the character string of each data included in the non-RDF data set 826 and the literal character string of each node included in the RDF data set 825 match. Then, the search unit 813 identifies a node having a literal that matches the character string of each data as a corresponding node, and generates a search node queue 827 including a plurality of corresponding nodes (step 1302).

  FIG. 14 shows an example of the search node queue 827. Corresponding nodes n1 to n4 are registered in the search node queue 827 of FIG.

  Next, the search unit 813 checks whether the search node queue 827 is empty (step 1303). When the search node queue 827 is not empty (step 1303, NO), the search unit 813 extracts one node from the search node queue 827 as a movement source node (step 1304) and performs node movement processing (step 1305).

  Next, the search unit 813 refers to the route list 828 to check whether or not there is a common node that directly or indirectly refers to all the corresponding nodes identified in step 1301 (step 1306). . When there is no common node that directly or indirectly refers to all the corresponding nodes (step 1306, NO), the search unit 813 repeats the processing from step 1303.

  On the other hand, when there is a common node that directly or indirectly refers to all the corresponding nodes (step 1306, YES), the search unit 813 generates a search result 829 including information indicating the common nodes detected so far. To do. Then, the output unit 814 outputs the search result 829 (step 1307). In this case, the search result 829 includes information indicating the common node and information indicating the corresponding node that the common node directly or indirectly refers to, for all the common nodes detected so far.

  When the search node queue 827 is empty (step 1303, YES), the node search device 801 performs the process of step 1307.

  FIG. 15 is a flowchart showing an example of the node move processing in step 1305 of FIG. First, the search unit 813 identifies, in the RDF data set 825, one of the nodes at both ends of the link that refers to the source node, which is not the source node, as the destination node (step 1501). Then, the search unit 813 determines whether to move the search position from the source node to the destination node (step 1502).

For example, the search unit 813 determines that there is a relationship between the ontology of the source node and the ontology of the destination node when either of the following condition (C1) or condition (C2) is satisfied. In this case, it is determined that the search position is moved from the source node to the destination node.
(C1) The co-occurrence statistic 823 for the link that refers to the source node and the link that refers to the destination node is larger than the predetermined value α.

For example, α is a value designated by the user and may be α = 0. When there is no link that refers to the destination node, the co-occurrence statistic 823 is set to a very large value. In this case, the maximum value of 64-bit floating point numbers may be used as the co-occurrence statistic 823.
(C2) The important word included in the label and comment of the link that refers to the source node overlaps with the important word included in the label and comment of the link that refers to the destination node.

  The important word included in the label and comment of the link is a word having the importance statistic 824 larger than the predetermined value β among the words included in the label and the comment. If at least one important word of the link that refers to the source node is the same as at least one important word of the link that refers to the destination node, it is determined that the important words of those links are duplicated. For example, β is a value specified by the user.

  On the other hand, when neither the condition (C1) nor the condition (C2) is satisfied, the search unit 813 determines that there is no relationship between the ontology of the source node and the ontology of the destination node. In this case, it is determined that the search position is not moved from the source node to the destination node.

  In addition, in the RDF data set 825, when a plurality of links refer to the movement source node, the search unit 813 determines whether or not to move the search position for each movement destination node of the end point of those links. .

  When a plurality of links refer to the movement-destination node, the search unit 813 satisfies the condition (C1) or the condition (C2) for all combinations of those links and the links referencing the movement-source node. Or not. Then, the search unit 813 determines to move the search position when the condition (C1) or the condition (C2) is satisfied for any combination.

  When moving the search position (step 1502, YES), the search unit 813 updates the route list 828 (step 1503) and detects a common node based on the updated route list 828 (step 1504). Then, the search unit 813 adds the destination node to the search node queue 827 (step 1505). The added destination node is extracted as a new source node in step 1304 of FIG.

  On the other hand, when the search position is not moved (step 1502, NO), the search unit 813 ends the process without updating the route list 828.

The route list 828 includes an entry indicating the route of each node detected by the node detection process. The entry indicating the route of the node m includes identification information of one or more nodes existing between any corresponding node and the node m in the RDF data set 825. If the node m is the corresponding node, the entry contains only the identification information of the node m. In step 1503, the search unit 813 updates the route list 828 in the following procedure, for example.
(P31) The search unit 813 acquires the entry E0 indicating the route of the source node from the route list 828.
(P32) The search unit 813 deletes the entry E0 from the route list 828. (P33) The search unit 813 adds the identification information of the movement destination node to the entry E0 and generates the entry E1 indicating the route of the movement destination node. To do. If there are multiple destination nodes, an entry E1 is created for each destination node.
(P34) The search unit 813 adds the entry E1 to the route list 828.
(P35) The search unit 813 checks the entries included in the route list 828, and if there are multiple entries indicating the route of the same node, deletes those entries from the route list 828. Then, the search unit 813 combines the deleted entries to generate a new entry E2 and adds it to the route list 828.

  FIG. 16 shows an example of paths included in the RDF data set 825. n0, n1, n00, n11, n000, n001, and n002 represent nodes, and arrows between the nodes represent links. n0 and n1 are corresponding nodes.

  FIG. 17 shows an example of the route list 828 showing the routes of FIG. FIG. 17A shows an example of the entry E0 acquired from the route list 828. The entry 1701 indicates the route of the source node n00, and p (n0, n00) represents a node existing on the route from the corresponding node n0 to the source node n00.

  FIG. 17B shows an example of the entry E1 generated from the entry E0 in FIG. 17A. In this case, the destination nodes that can be moved from the source node n00 are the nodes n000 to n002. Among them, when the movement destination node n001 and the movement destination node n002 are determined as the movement destinations of the search position, the entry 1702 and the entry 1703 indicating the routes of these movement destination nodes are generated as the entry E1.

  P (n0, n00, n001) of the entry 1702 represents a node existing on the path from the corresponding node n0 to the destination node n001. Further, p (n0, n00, n002) of the entry 1703 represents a node existing on the route from the corresponding node n0 to the movement-destination node n002.

  FIG. 18 shows an example of the route list 828 including the common node. FIG. 18A shows an example of the entry E0 acquired from the route list 828. The entry 1801 shows the route of the source node n00, and the entry 1802 shows the route of the source node n11. The p (n0, n00) of the entry 1801 represents a node existing on the path from the corresponding node n0 to the source node n00, and the p (n1, n11) of the entry 1802 is from the corresponding node n1 to the source node n11. Represents a node existing on the path of.

  FIG. 18B shows an example of the entry E1 generated from the entry E0 of FIG. 18A. Among the destination nodes that can be moved from the source node n00, if the destination node n000 and the destination node n001 are determined as the destinations of the search positions, the entry 1803 indicating the route of these destination nodes is set as the entry E1. And an entry 1804 is created.

  The entry 1803 p (n0, n00, n000) represents a node existing on the route from the corresponding node n0 to the destination node n000. Further, p (n0, n00, n001) of the entry 1804 represents a node existing on the route from the corresponding node n0 to the destination node n001.

  Furthermore, when the destination node n001 that can be moved from the source node n11 is determined as the destination of the search position, an entry 1805 indicating the route of the destination node n001 is generated as the entry E1. Entry 1805 p (n1, n11, n001) represents a node existing on the path from the corresponding node n1 to the destination node n001.

  18C shows an example of a new entry E2 generated from the entry 1804 and the entry 1805 of FIG. 18B in the procedure (P35). The entry 1804 and the entry 1805 indicate the route of the same node n001, and thus are deleted from the route list 828. Then, the entry 1806 is generated by combining the entry 1804 and the entry 1805.

  The entry 1806 includes identification information of the node n001, p (n0, n00, n001) indicating the route from the corresponding node n0 to the node n001, and p (n1, n11, n001) indicating the route from the corresponding node n1 to the node n001. ) And are included.

  Therefore, the search unit 813 can detect that the route that follows the link from the corresponding node n0 and the route that follows the link from the corresponding node n1 intersect at the node n001 by checking the entry 1806 in step 1504. As a result, the node n001 is detected as the common node. In this case, the search unit 813 generates a search result 829 including the identification information (n001, n0, n1) of the common node n001, the corresponding node n0, and the corresponding node n1.

  The configurations of the node search device 601 of FIG. 6 and the node search device 801 of FIG. 8 are merely examples, and some of the constituent elements may be omitted or changed depending on the use or condition of the node search device. For example, in the node search device 801 of FIG. 8, when the co-occurrence statistic 823 and the importance statistic 824 are calculated by an external device, or when the co-occurrence statistic 823 and the importance statistic 824 are not used, The calculation unit 812 can be omitted.

  The flowcharts of FIGS. 7, 12, 13, and 15 are merely examples, and some processing may be omitted or changed depending on the configuration or conditions of the node search device. For example, in the node search process of FIG. 12, when the co-occurrence statistic 823 and the importance statistic 824 are calculated by an external device, the processes of step 1201 and step 1202 can be omitted. Even when the co-occurrence statistic 823 and the importance statistic 824 are not used in the node search process of FIG. 12, the processes of step 1201 and step 1202 can be omitted.

  In step 1502 of FIG. 15, the search unit 813 may determine whether to move the search position using only one of the condition (C1) and the condition (C2), and the search may be performed using another condition. It may be determined whether to move the position. For example, the user can store in the storage unit 811 a list indicating a combination of two links having a relationship between ontologies. In this case, the search unit 813 checks whether a combination of a link that refers to the source node and a link that refers to the destination node is included in the list to determine whether to move the search position. To decide.

  The RDF datasets illustrated in FIGS. 1 to 3, 4B, 5 and 9 to 11 are merely examples, and the RDF dataset is used for description and information described by RDF. It changes according to the ontology. The non-RDF data set shown in FIG. 4A is merely an example, and the non-RDF data set changes according to the information to be described.

  The search node queue 827 shown in FIG. 14 is merely an example, and the corresponding nodes included in the search node queue 827 change according to the RDF data set 825 and the non-RDF data set 826. The route shown in FIG. 16 and the route list 828 shown in FIGS. 17 and 18 are merely examples, and the route and the route list 828 change according to the RDF data set 825 and the non-RDF data set 826.

  Expressions (1) to (15) are merely examples, and the search unit 813 may calculate the co-occurrence statistic 823 and the importance statistic 824 using another calculation expression.

  FIG. 19 shows a configuration example of an information processing device (computer) used as the node search device 601 of FIG. 6 and the node search device 801 of FIG. The information processing apparatus of FIG. 19 includes a CPU (Central Processing Unit) 1901, a memory 1902, an input device 1903, an output device 1904, an auxiliary storage device 1905, a medium drive device 1906, and a network connection device 1907. These components are connected to each other by a bus 1908.

  The memory 1902 is, for example, a semiconductor memory such as a ROM (Read Only Memory), a RAM (Random Access Memory), and a flash memory, and stores programs and data used for processing. The memory 1902 can be used as the storage unit 611 in FIG. 6 or the storage unit 811 in FIG.

  The CPU 1901 (processor) operates as the search unit 612 in FIG. 6, the calculation unit 812 in FIG. 8, and the search unit 813 by executing the program using the memory 1902, for example.

  The input device 1903 is, for example, a keyboard, a pointing device, or the like, and is used to input an instruction or information from an operator or a user. The output device 1904 is, for example, a display device, a printer, a speaker, or the like, and is used for inquiring or instructing an operator or a user and outputting a processing result. The output device 1904 can be used as the output unit 613 of FIG. 6 or the output unit 814 of FIG. The processing result may be the search result 829.

  The auxiliary storage device 1905 is, for example, a magnetic disk device, an optical disk device, a magneto-optical disk device, a tape device, or the like. The auxiliary storage device 1905 may be a hard disk drive or a flash memory. The information processing apparatus can store the program and data in the auxiliary storage device 1905 and load them into the memory 1902 for use. The auxiliary storage device 1905 can be used as the storage unit 611 of FIG. 6 or the storage unit 811 of FIG.

  The medium driving device 1906 drives a portable recording medium 1909 to access the recorded contents. The portable recording medium 1909 is a memory device, flexible disk, optical disk, magneto-optical disk, or the like. The portable recording medium 1909 may be a CD-ROM (Compact Disk Read Only Memory), a DVD (Digital Versatile Disk), a USB (Universal Serial Bus) memory, or the like. The operator or the user can store the program and data in the portable recording medium 1909 and load them into the memory 1902 for use.

  As described above, a computer-readable recording medium that stores programs and data used for processing is a physical (non-transitory) recording medium such as the memory 1902, the auxiliary storage device 1905, or the portable recording medium 1909. It is a medium.

  The network connection device 1907 is a communication interface circuit that is connected to a communication network such as a LAN (Local Area Network) and a WAN (Wide Area Network) and performs data conversion accompanying communication. The information processing device can receive a program and data from an external device via the network connection device 1907, load them into the memory 1902, and use them. The network connection device 1907 can be used as the output unit 613 of FIG. 6 or the output unit 814 of FIG.

  The information processing device can also receive the RDF data set 825, the non-RDF data set 826, and the processing request from the user terminal via the network connection device 1907, and can also transmit the search result 829 to the user terminal.

  Note that the information processing device does not need to include all the constituent elements of FIG. 19, and it is possible to omit some of the constituent elements according to the use or the conditions. For example, when the information processing device receives a processing request from the user terminal, the input device 1903 and the output device 1904 may be omitted. When the portable recording medium 1909 or the communication network is not used, the medium driving device 1906 or the network connection device 1907 may be omitted.

  Although the disclosed embodiments and their advantages have been described in detail, those skilled in the art can make various changes, additions, and omissions without departing from the scope of the invention explicitly described in the claims. Let's do it.

Regarding the embodiment described with reference to FIGS. 4 to 19, the following supplementary notes are further disclosed.
(Appendix 1)
A node search method executed by a computer, comprising:
The computer is
From the nodes included in the first data set described based on the multiple ontologies, the nodes corresponding to each of the multiple data items included in the second data set are specified,
On a route that follows a link included in the first data set from each of the identified nodes, the relationship between the ontology of the source node and the ontology of the destination node is determined,
A first route that follows a link from a first node of the plurality of nodes by tracing a link between the source node and the destination node based on a result of determining the relevance; A common node intersecting with a second route that follows the link from the second node among the nodes
Outputting a search result including information indicating the common node, information indicating the first node, and information indicating the second node,
A node search method characterized by the above.
(Appendix 2)
When the computer determines that there is a relationship between the ontology of the source node and the ontology of the destination node, the computer follows a link between the source node and the destination node, The node is set as a new source node, and the search on the route including the new source node is continued, and if there is no relationship between the ontology of the source node and the ontology of the destination node. If the determination is made, the search on the route including the source node is aborted, and the node search method according to note 1.
(Appendix 3)
The computer obtains the number of appearances of the two links appearing simultaneously for a combination of two links included in each of the plurality of data sets, and calculates the co-occurrence statistic for the two links using the obtained number of appearances. Between the ontology of the source node and the ontology of the destination node based on the co-occurrence statistic for the first link that refers to the source node and the second link that refers to the destination node. The node search method according to supplementary note 1 or 2, wherein the relevance is determined.
(Appendix 4)
The computer, from the ontology that defines each link included in the plurality of data sets, extracts the words included in the label and comment of each link, and obtains the number of appearances of each word that appears in the label and comment of each link, An importance statistic indicating the importance of each word is calculated using the obtained number of appearances, and a first importance indicating the importance of each word included in the label and comment of the first link that refers to the source node. The ontology of the source node and the destination node are calculated using a statistic and a second importance statistic indicating the importance of each word included in the label and comment of the second link that references the destination node. 3. The node search method according to appendix 1 or 2, characterized in that the relevance to the above ontology is determined.
(Appendix 5)
From the nodes included in the first data set described based on the multiple ontologies, the nodes corresponding to each of the multiple data items included in the second data set are specified,
On a route that follows a link included in the first data set from each of the identified nodes, the relationship between the ontology of the source node and the ontology of the destination node is determined,
A first route that follows a link from a first node of the plurality of nodes by tracing a link between the source node and the destination node based on a result of determining the relevance; A common node intersecting with a second route that follows the link from the second node among the nodes
Outputting a search result including information indicating the common node, information indicating the first node, and information indicating the second node,
A node search program that causes a computer to execute processing.
(Appendix 6)
When the computer determines that there is a relationship between the ontology of the source node and the ontology of the destination node, the computer follows a link between the source node and the destination node, The node is set as a new source node, and the search on the route including the new source node is continued, and if there is no relationship between the ontology of the source node and the ontology of the destination node. If the determination is made, the search on the route including the source node is aborted, and the node search program according to note 5.
(Appendix 7)
The computer obtains the number of appearances of the two links appearing simultaneously for a combination of two links included in each of the plurality of data sets, and calculates the co-occurrence statistic for the two links using the obtained number of appearances Between the ontology of the source node and the ontology of the destination node based on the co-occurrence statistic for the first link that refers to the source node and the second link that refers to the destination node. 7. The node search program according to supplementary note 5 or 6, which determines the relevance.
(Appendix 8)
The computer, from the ontology that defines each link included in the plurality of data sets, extracts the words included in the label and comment of each link, and obtains the number of appearances of each word that appears in the label and comment of each link, An importance statistic indicating the importance of each word is calculated using the obtained number of appearances, and a first importance indicating the importance of each word included in the label and comment of the first link that refers to the source node. The ontology of the source node and the destination node are calculated using a statistic and a second importance statistic indicating the importance of each word included in the label and comment of the second link that references the destination node. 7. The node search program according to appendix 5 or 6, wherein the relevance to the ontology is determined.

301, 311, 312 Data sets 321 to 326 Ontology 601, 80 1 Node search device 611, 811 Storage unit 612, 813 Search unit 613, 814 Output unit 621 First data set 622 Second data set 812 Calculation unit 821 RDF data set group 822 Ontology group 823 Co-occurrence statistic 824 Importance statistic 825 RDF data set 826 Non-RDF data set 827 Search node queue 828 Route list 829 Search result 1701-1703, 1801-1806 Entry 1901 CPU
1902 Memory 1903 Input Device 1904 Output Device 1905 Auxiliary Storage Device 1906 Medium Drive Device 1907 Network Connection Device 1908 Bus 1909 Portable Recording Medium

Claims (5)

A node search method executed by a computer, comprising:
The computer is
From the nodes included in the first data set described based on the multiple ontologies, the nodes corresponding to each of the multiple data items included in the second data set are specified,
On a route that follows a link included in the first data set from each of the identified nodes, the relationship between the ontology of the source node and the ontology of the destination node is determined,
A first route that follows a link from a first node of the plurality of nodes by tracing a link between the source node and the destination node based on a result of determining the relevance; A common node intersecting with a second route that follows the link from the second node among the nodes
Outputting a search result including information indicating the common node, information indicating the first node, and information indicating the second node,
A node search method characterized by the above.   When the computer determines that there is a relationship between the ontology of the source node and the ontology of the destination node, the computer follows a link between the source node and the destination node, If the node is set as a new source node and the search on the route including the new source node is continued, and there is no relationship between the ontology of the source node and the ontology of the destination node The node search method according to claim 1, wherein when the determination is made, the search on the route including the source node is terminated.   The computer obtains the number of appearances of the two links appearing simultaneously for a combination of two links included in each of the plurality of data sets, and calculates the co-occurrence statistic for the two links using the obtained number of appearances. Between the ontology of the source node and the ontology of the destination node based on the co-occurrence statistic for the first link that refers to the source node and the second link that refers to the destination node. 3. The node search method according to claim 1, wherein the relevance is determined.   The computer, from the ontology that defines each link included in the plurality of data sets, extracts the words included in the label and comment of each link, and obtains the number of appearances of each word that appears in the label and comment of each link, An importance statistic indicating the importance of each word is calculated using the obtained number of appearances, and a first importance indicating the importance of each word included in the label and comment of the first link that refers to the source node. The ontology of the source node and the destination node are calculated using a statistic and a second importance statistic indicating the importance of each word included in the label and comment of the second link that references the destination node. 3. The node search method according to claim 1, further comprising determining the relevance to the ontology of. From the nodes included in the first data set described based on the multiple ontologies, the nodes corresponding to each of the multiple data items included in the second data set are specified,
On a route that follows a link included in the first data set from each of the identified nodes, the relationship between the ontology of the source node and the ontology of the destination node is determined,
A first route that follows a link from a first node of the plurality of nodes by tracing a link between the source node and the destination node based on a result of determining the relevance; A common node intersecting with a second route that follows the link from the second node among the nodes
Outputting a search result including information indicating the common node, information indicating the first node, and information indicating the second node,
A node search program that causes a computer to execute processing.

Images